Theta (θ) oscillations are one of the characteristic local field potentials (LFPs) in the hippocampus that emerge during spatial navigation, exploratory sniffing, and rapid eye movement sleep. LFPs are thought to summarize multineuronal events, including synaptic currents and action potentials. However, no in vivo study to date has directly interrelated θ oscillations with the membrane potentials (Vm) of multiple neurons, and it remains unclear whether LFPs can be predicted from multineuronal Vms. Here, we simultaneously patch-clamp up to three CA1 pyramidal neurons in awake or anesthetized mice and find that the temporal evolution of the power and frequency of θ oscillations in Vms (θVms) are weakly but significantly correlate with LFP θ oscillations (θLFP) such that a deep neural network could predict the θLFP waveforms based on the θVm traces of three neurons. Therefore, individual neurons are loosely interdependent to ensure freedom of activity, but they partially share information to collectively produce θLFP.